42,000+ OpenClaw AI agent servers exposed on internet without authentication
Security researchers discovered approximately 42,000 OpenClaw AI agent servers exposed publicly on the internet without authentication across 80+ countries. The systems were vulnerable to remote code execution, credential theft, and potential network compromise due to default insecure configurations and an unvetted third-party skill marketplace.
What happened
Security researchers discovered around 42,000 instances of OpenClaw AI agent servers directly exposed on the open internet across at least 80 countries. Many instances were configured with no authentication and vulnerable to remote code execution, allowing unauthorised users to control systems, access API keys, chat histories, and credentials.
OpenClaw's control panels were bound to all network interfaces by default without mandatory authentication. The third-party 'skill' marketplace ClawHub contained hundreds of malicious skills that could steal data or deliver malware.
Reported impact
The exposure demonstrates that compromises to AI agents constitute physical actions on data or finances, not merely text hallucinations. A compromised personal agent could become a backdoor into an entire corporate network, bypassing traditional firewalls.
Index analysis
The incident stems from poor security defaults: administrative interfaces bound to all network interfaces without authentication, lack of mandatory authentication or least-privilege checks, and absence of robust review or security sandboxing in the skill marketplace.
SecurityScorecard's reconnaissance identified 15.2K exposed instances vulnerable to remote code execution, with 35.4% of observed deployments flagged as vulnerable. Three high-severity CVEs with public exploit code were available, and 53.3K exposed instances correlated with prior breach activity.
Governance controls
The incident highlights the need for security-by-default standards for AI agents, including mandatory 'human-in-the-loop' requirements for high-risk actions like deleting data or moving funds. Regulators may need to hold developers accountable for the safety of automated plugin marketplaces.
Classification
Sources and evidence
This record was researched and written by the Index, based on the following original reporting.